Many-Shot In-Context Learning for Molecular Inverse Design
Abstract
Large Language Models (LLMs) have demonstrated great performance in few-shot In-Context Learning (ICL) for a variety of generative and discriminative chemical design tasks. The newly expanded context windows of LLMs can further improve ICL capabilities for molecular inverse design and lead optimization. To take full advantage of these capabilities we developed a new semi-supervised learning method that overcomes the lack of experimental data available for many-shot ICL. Our approach involves iterative inclusion of LLM generated molecules with high predicted performance, along with experimental data. We further integrated our method in a multi-modal LLM which allows for the interactive modification of generated molecular structures using text instructions. As we show, the new method greatly improves upon existing ICL methods for molecular design while being accessible and easy to use for scientists.
Cite
Text
Moayedpour et al. "Many-Shot In-Context Learning for Molecular Inverse Design." ICML 2024 Workshops: AI4Science, 2024.Markdown
[Moayedpour et al. "Many-Shot In-Context Learning for Molecular Inverse Design." ICML 2024 Workshops: AI4Science, 2024.](https://mlanthology.org/icmlw/2024/moayedpour2024icmlw-manyshot/)BibTeX
@inproceedings{moayedpour2024icmlw-manyshot,
title = {{Many-Shot In-Context Learning for Molecular Inverse Design}},
author = {Moayedpour, Saeed and Corrochano-Navarro, Alejandro and Sahneh, Faryad and Koetter, Alexander and Vymětal, Jiří and Anele, Lorenzo Kogler and Mas, Pablo and Jangjoo, Yasser and Li, Sizhen and Bailey, Michael and Bianciotto, Marc and Matter, Hans and Grebner, Christoph and Hessler, Gerhard and Bar-Joseph, Ziv and Jager, Sven},
booktitle = {ICML 2024 Workshops: AI4Science},
year = {2024},
url = {https://mlanthology.org/icmlw/2024/moayedpour2024icmlw-manyshot/}
}